Regression relations for bankfull stream characteristics based on drainage area (often called 'regional curves') are used in natural stream channel design to verify field determinations of bankfull discharge and stream channel characteristics. Bankfull stream characteristics were assessed for stream reaches at 41 streamflow-gaging stations in the Valley and Ridge Physiographic Province in Maryland, Virginia, and West Virginia. Data collected included bankfull cross-sectional geometry, flood plain geometry, and longitudinal profile data. In addition, particle-size distributions of streambed material were determined and data on basin characteristics were compiled for each reach. Regional curves were developed for bankfull cross-sectional area, width, and discharge with R2 values of 0.95, 0.89, 0.87, and 0.91, respectively. Examination of the regional curves residuals indicates that there is more variability in bankfull cross-sectional area, width, and discharge for smaller streams than for larger streams. In contrast, there is more variability for bankfull mean depth for larger streams than for smaller streams. Geographic analysis of regional curve residuals indicated that there were no further subdivisions within the Valley and Ridge Physiographic Province in the three-state study area for which individual sets of regional curves should be developed. In addition, two separate sets of regional curves were developed with data from the 41 sites to examine potential differences in the relations between the southern (n = 9) and central (n = 32) sections of the province. There were differences in slope and intercept between the two bankfull discharge test relations and a difference in intercept for the width test relations at the 95-percent confidence level. However, the results of this analysis were inconclusive and therefore one set of regional curves for the study area is presented in this report. The regional curves were compared to regression models developed from similar data collected in the Pennsylvania and Maryland portions of the province. No statistical difference in the slope or intercept of regression lines of the three data sets was detected for any of the four bankfull parameters at the 95-percent confidence level. Basin characteristics such as percentage of basin forested (percent forested) and percentage of basin underlain by carbonate bedrock (percent carbonate) were analyzed to evaluate variability among regression points. Multivariate regression relations including explanatory terms for percent carbonate and drainage area produced higher R2 values than the regional curves for bankfull cross-sectional area (R2 = 0.95), bankfull width (R2 = 0.92), and bankfull discharge (R2 = 0.93). There was no improvement for the bankfull mean depth relation from adding the additional term. Inclusion of the other basin characteristics in multivariate relations did not improve the regression models. Regression models developed for the 1.5-year discharge for all streamflow-gaging stations with peak discharge data throughout Virginia (n = 486) and throughout the Valley and Ridge Physiographic Province in Virginia (n = 147) were compared to the regional curve relating bankfull discharge to drainage area. A similar trend in decreasing variability with increasing drainage area was observed for the 1.5-year discharge for all stations in Virginia . This indicates that the change in variability observed in the discharge regional curve likely would exist with a larger data set. There was no statistical difference at the 95-percent confidence level between regression relations for the southern section of the province (n = 40) and the central section (n = 107). This finding supports maintaining only one set of regional curves for the study area. Not all of the variability in the regional curves is explained by drainage area alone. Causes of the remaining variability likely vary among study sites. Users of the regional curves de